The primary goals of treatment after ACL injury are thereduction or elimination of knee instability, the restorationof lost function, and the prevention of long-term jointdegeneration. Although modern ACL reconstruction proce-dures are arguably successful for meeting the first 2 goals,long-term joint health after ACL injury remains a concern.

The ACL injury has long been associated with a highincidence of degenerative knee osteoarthritis (OA).14,25,50

Animal models have shown a clear connection betweenmechanical instability and progressive OA.4,31,32 Thus, it islogical to assume that if ACL reconstruction surgery

restores knee stability, it should reduce the risk of subse-quent OA. However, a similar or higher incidence of kneeOA has been reported in ACL-deficient individuals whounderwent ligament reconstruction as compared to thosewho did not.11,15 Despite criticisms of these studies (retro-spective, uncontrolled, possible selection bias betweenpatients who choose surgical reconstruction vs those whodo not), to date there is no clear evidence of a long-termprotective effect from the procedure. Although other fac-tors may contribute to OA development in the recon-structed joint (such as possible damage to other tissues atthe time of injury), failure of the reconstruction to restorenormal knee kinematics could be an important contributorto progressive cartilage degeneration.

The effectiveness of ACL reconstruction for restoringnormal knee function has been difficult to assess. A varietyof radiographic and mechanical measurement methodshave been employed to assess static knee stability in vivoby measuring the anterior displacement of the tibia rela-tive to the femur resulting from a specified anterior force

Abnormal Rotational Knee Motion DuringRunning After Anterior Cruciate LigamentReconstructionScott Tashman,*† PhD, David Collon,‡ MD, Kyle Anderson,‡ MD, Patricia Kolowich,‡ MD,and William Anderst,† MSFrom the †Bone and Joint Center and the ‡Department of Orthopaedic Surgery, Henry Ford Health System, Detroit, Michigan

Background: The effectiveness of anterior cruciate ligament reconstruction for restoring normal knee kinematics is largelyunknown, particularly during sports movements generating large, rapidly applied forces.

Hypothesis: Under dynamic in vivo loading, significant differences in 3-dimensional kinematics exist between anterior cruciateligament–reconstructed knees and the contralateral, uninjured knees.

Study Design: Prospective, in vivo laboratory study.

Methods: Kinematics of anterior cruciate ligament–reconstructed and contralateral (uninjured) knees were evaluated for 6 sub-jects during downhill running 4 to 12 months after anterior cruciate ligament reconstruction, using a 250 frame/s stereoradi-ographic system. Anatomical reference axes were determined from computed tomography scans. Kinematic differencesbetween the uninjured and reconstructed limbs were evaluated with a repeated-measures analysis of variance.

Results: Anterior tibial translation was similar for the reconstructed and uninjured limbs. However, reconstructed knees weremore externally rotated on average by 3.8 ± 2.3° across all subjects and time points (P = .0011). Reconstructed knees were alsomore adducted, by an average of 2.8 ± 1.6° (P = .0091). Although differences were small, they were consistent in all subjects.

Conclusions: Anterior cruciate ligament reconstruction failed to restore normal rotational knee kinematics during dynamic loading.

Clinical Relevance: Although further study is required, these abnormal motions may contribute to long-term joint degenerationassociated with anterior cruciate ligament injury/reconstruction.

Keywords: knee; anterior cruciate ligament (ACL); kinematics; reconstruction; running

975

DOI = 10.1177/0363546503261709

*Address correspondence and reprint requests to Scott Tashman,PhD, Bone and Joint Center, ER2015, Henry Ford Health System, Detroit,MI 48202 (e-mail: tashman@bjc.hfh.edu).

No author or related institution has received financial benefit fromresearch in this study.

The American Journal of Sports Medicine, Vol. 32, No. 4DOI: 10.1177/0363546503261709© 2004 American Orthopaedic Society for Sports Medicine

976 Tashman et al The American Journal of Sports Medicine

applied to the tibia with the femur held in a fixed position.These studies have shown static instability (typically 5-10mm) following ACL loss as well as improvement to near-normal levels following ACL reconstruction.10,21,26,42,48

Three-dimensional (3D) radiographic techniques have alsoshown increased rotational laxity (internal/external tibialrotation) in ACL-deficient knees.20 However, static sta-bility measures have not correlated well with any knownmeasure of functional outcome for ACL-injured subjectsbefore or after reconstruction.3,9,17,39,43 In addition, thereare reports of individuals with statically unstable kneesperforming asymptomatically at high levels of athleticactivity,28,30 suggesting that static stability may not be aprerequisite for good knee function.

Why is static stability not predictive of outcome? Duringfunctional activities, the knee is subjected to a combinationof dynamic physical forces (gravitational, inertial, and con-tact) and active muscular forces.27,37 The combination ofthese forces and the constraints imposed by passive struc-tures (articular surface geometry, ligaments, etc) lead to acomplex motion path, with rotations and translations in all3 planes of movement.23 Static, 1-dimensional testing can-not predict the behavior of the reconstructed joint underrealistic loading conditions. Thus, dynamic assessment of3D skeletal kinematics is necessary to assess the effective-ness of ACL reconstruction surgery for restoring normalknee function.

Accurate, reliable 3D knee-motion measurement is prob-lematic with commonly used motion analysis techniques.Conventional motion-measurement methods, employingeither optoelectronic or video-based systems, rely on mark-ers attached to the skin to estimate joint motion. Studiesusing implanted bone pins have shown that markersaffixed to the skin shift relative to underlying bone by asmuch as 30 mm, particularly during rapid movements oractivities involving impact such as heelstrike.18,24,35 Thismarker-tracking error varies with bone, marker position,activity, and joint angle and is often correlated with themovement, complicating efforts to develop algorithms forerror correction.6 Techniques have been developed combin-ing large numbers of skin markers with algorithms todetect violations of the rigid body assumption and modelsoft tissue deformation.1,2,7,44 A single-subject validationstudy demonstrated that this approach can track tibialmotion well during slow, nonimpact movements (low-speed10 cm step-up; average error = 0.8 mm, peak error =approximately 2.6 mm).1 However, the performance fortracking the femur, where errors are likely to be higherbecause of greater soft tissue thickness between markersand bone, has not been reported. Validation studies duringfaster movements including impact events (eg, gait, run-ning, jumping), where larger magnitude and more complexskin deformation occurs, are not available.

It is possible to avoid skin motion artifacts by directlymeasuring bone movement, either by physically attachingmeasurement devices to the bone or via medical-imagingtechniques. Accurate 3D in vivo kinematics studies havebeen performed using external marker arrays rigidly fixedto bone.23 However, the risks and discomfort associated

with this technique have prevented widespread use andmade serial studies difficult. Dynamic MRI and CT meth-ods show promise36,41 but are limited by low frame ratesand environments too restrictive for most dynamic,weightbearing activities.

Radiographic imaging offers a minimally invasive alter-native for knee-motion assessment. Conventional fluo-roscopy permits direct visualization of bone motion but islimited to 2-dimensional assessment and is prone to errorsdue to parallax and motion blur. Biplane or stereo radio-graphic imaging enables accurate quantitative 3D motionassessment as well as direct visualization of bone motion.Use of biplane radiographic film methods (radiographicstereophotogrammetric analysis [RSA]) for 3D studies ofstatic bone position has been well established,21,38 withprecision reported in the ±10 to 250 µm range.19 We havedeveloped a dynamic RSA system for accurate assessmentof dynamic joint motion. This system (described briefly inthe Methods section below and in detail elsewhere45) iswell suited for dynamic knee-motion measurement, with3D accuracy in the order of ±0.1 mm and rates up to 1000frames/s.

The goal of this study was to use this high-speed biplaneradiography system to assess the 3D kinematics of healthyand ACL-reconstructed knees during dynamic, functionalactivities. It was hypothesized that accurate 3D measure-ment would reveal differences in dynamic joint motionbetween reconstructed and healthy joints. If true, thesefindings might help to explain why reconstruction is notprotective against long-term joint degeneration associatedwith ACL injury. With this information, it may become pos-sible to design improved reconstruction techniques thatrestore more normal joint motion.

MATERIALS AND METHODS

Subjects undergoing unilateral, primary arthroscopic ACLreconstruction and who were between the ages of 16 and50 years were recruited for the study. Exclusion criteriaincluded any prior significant injury to the contralaterallimb as well as significant damage to other knee structuresin the ACL-injured limb (subjects with minor meniscaltears, requiring removal of no more than one third of theradial width of the meniscus, were not excluded). Allpatients who were not excluded by these criteria and whounderwent ACL reconstruction by 1 of the 3 participatingsurgeons were asked to participate in the study. Informedconsent was obtained from all enrolled subjects, and theprotocol was approved by the appropriate institutionalreview board for human subject research. During the ACLreconstruction procedure, tantalum spheres (1.6-mmdiameter) were inserted into the distal femur and proximaltibia of both limbs using a cannulated drill. These markersprovided high-accuracy radiographic targets for RSA andhave been used similarly in many human studies with noadverse reactions reported.19 Three noncollinear markerswere inserted into each bone to enable full 6 degrees offreedom tracking. The goal of marker placement was to

Vol. 32, No. 4, 2004 Abnormal Rotational Knee Motion During Running 977

ensure a minimum 2-cm separation between markers andto avoid overlap between markers or with any implantedhardware in the radiographic views. No effort was made toplace the markers at specific anatomical landmarks.

Kinematic testing was performed as soon as possibleafter the subject completed his or her rehabilitation pro-gram and was cleared by the surgeon for return to lightsports activities. This time period was typically 4 to 6months but was 12 months for 1 subject (the first in thestudy; data collected at 4 months for this subject wereunusable due to a technical problem). Within 2 weeks ofthe kinematic testing, static knee laxity was assessed(using a KT-1000 arthrometer at 89 N anterior force) forcomparison with the dynamic measurements.

Subjective knee function was assessed with the Cincin-nati Knee Ligament Rating System.29 The 4 reported val-ues in the Cincinnati Knee Ligament Rating System are(1) current sports activity level, (2) ability to participate insports, (3) functional assessment rating scale, and (4) rateof pain, swelling, and giving way. All were based on a 0% to100% scoring system, with 100% being maximum func-tion/minimum pain. In a comparison of several systems(applied to 65 patients at a mean of 35 months after ACLreconstruction), the Cincinnati system was recommendedbecause it was “the most elaborate and detailed assess-ment of activity-related symptoms and function,” and it“most precisely defines outcome in athletically activepatients.”40

Moderate-speed downhill running was selected as themovement activity for study. This task was selectedbecause it is more stressful on the ACL than level-groundrunning.22 In addition to placing greater mechanicaldemands on the knee than walking, running also elimi-nates the double-support phase and reduces the effects ofcompensation from the contralateral limb. It can be per-formed in a controlled, repeatable fashion within the labo-ratory environment and is unlikely to put the individual atsignificant risk for injury. Downhill running has been usedpreviously to assess performance in ACL-deficient individ-uals.47 A moderately slow running speed (“jog”) was se-lected to ensure that all subjects would be able to performthe task. Testing was performed at 2.5 m/s on a standardtreadmill (Landice L8, 46 × 152 cm belt, Landice Inc,Randolph, NJ) with the rear supports elevated 25 cm toprovide a 10° downward slope (Figure 1). For each trial,kinematic data were collected from shortly before foot-strike through midstance for 1 step of the test leg (approx-imately 0.5 seconds duration), using an electronic timersystem and an accelerometer strapped to the shank todetect footstrike.

Knee kinematics were assessed with dynamic RSA. TheRSA is a technique for determining 3D kinematic informa-tion from stereo-pair radiographic images of musculoskele-tal tissue with implanted high-contrast markers.38 Thismethod has been successfully employed to determine staticor slow-moving 3D behavior of a variety of human jointsand joint structures, using pairs of plain film radio-graphs.19,38 Similar techniques were employed for thestudy of dynamic knee movement, replacing radiographic

film with high-speed digital imaging. The dynamic RSAsystem consisted of 2 gantries (each containing a 150 kWx-ray source, 30-cm image intensifier, and 250 frame/s digi-tal video system) configured to provide 2 beams parallel tothe ground, with an interbeam angle of 60°. This providedsufficient open space to incorporate a standard treadmilland a 3D-imaging area (where the beams intersect)approximately 45 cm wide by 30 cm long by 25 cm high(Figure 1). This system is capable of tracking implantedmarkers with an accuracy of approximately ±0.1 mm, aspreviously described.45

Two-dimensional marker positions were automaticallydetermined for each pair of digital images and passed to acommercial 3D-tracking software package (Eva, MotionAnalysis Corp, Santa Rosa, Calif) for calibration, tracking,and 3D reconstruction. The resulting 3D marker coordi-nates were smoothed using a fourth-order, zero-lagButterworth low-pass filter with a cutoff frequency of 25Hz (selected by residual analysis of several trials49).

Transformations between implanted marker-based coor-dinates and anatomical axes/landmarks were determinedfrom CT (Figure 2), as previously described.45 Rotations ofthe tibia relative to the femur were calculated using body-fixed axes in the order (flexion/extension, adduction/abduc-tion, internal/external rotation) corresponding to the rota-tional component of the joint coordinate system originallydescribed by Grood and Suntay.16 Translations for thereconstructed joint were determined relative to ACL graftorigin/insertion, estimated by identifying the center of thetibial and femoral tunnels at the joint surface of 3D bonemodels generated from subject-specific CT scans. To mini-mize side-to-side variability, the same locations were usedfor the uninjured limb by using a mirror image of thereconstructed limb CT and coregistering it with the unin-jured limb CT. Displacements of the tibia relative to thefemur were measured from ACL origin to insertion and

Figure 1. High-speed biplane radiographic system, config-ured for downhill treadmill running. Images are acquired simul-taneously at 250 frames/s for the 2 views (60° separation).

978 Tashman et al The American Journal of Sports Medicine

expressed in an orthogonal anatomical coordinate systemfixed to the tibia. Functional ACL-graft length was esti-mated as the magnitude of the origin-to-insertion vector.

Rotation and displacement curves were ensemble aver-aged over all valid trials (typically 3) for each test session.Kinematic variables (3 rotations, 3 translations, ACL-graftlength) were extracted at 0.0, 0.025, 0.05, 0.075, and 0.10seconds after footstrike for statistical analysis. Within-subject differences between the intact and reconstructedlimbs were evaluated via repeated-measures analysis ofvariance (significance set at P = .05).

RESULTS

A summary of subject characteristics is provided in Table 1.The subjects represented a heterogeneous mix of ages, sex,

graft type, time from injury, and meniscal health. Staticstability was restored in all subjects, with a mean side-to-side difference in KT-1000 arthrometer measurements of+1.2 mm (range, –1 to 2.5 mm). All subjects were able toperform the downhill-running task without difficulty, withno obvious limping or asymmetry. There were no statisti-cally significant correlations between age, sex, graft type,injury/surgery timing, static laxity, or functional scoresand any of the kinematic measures.

Joint kinematics for a typical subject are shown inFigures 3 and 4; group ensemble averages are provided inFigures 5 and 6. Knee flexion at footstrike ranged frommild hyperextension to nearly 20° of flexion in the unin-jured (intact) limbs. Although differences between intactand reconstructed limbs were noted in several subjects (eg,Figure 3A) and there was a slight trend toward loss ofextension in reconstructed knees (Figure 5A), there wereno statistically significant differences in flexion at anytime point (P = .71). Variability was highest in this plane,with mean across-subject standard deviations of 13° in theintact limbs and 8° in the reconstructed limbs.

Significant differences between limbs were identified inthe other 2 rotational axes (Figures 5B and C). Recon-structed knees were more externally rotated by 3.8 ± 2.3°(mean ± SD) across all subjects and time points (P = .0011).Reconstructed knees were also more adducted (by a meanof 2.8 ± 1.6°; P = .0091). Although these differences weresmall, they were observed consistently in all 6 subjects. Asshown in Figure 5, the shapes of the motion curves weresimilar for both limbs, but there were consistent baselineshifts toward external rotation and adduction on the recon-structed side. Across-subject standard deviations were lessthan 3° at all time points.

No significant differences were found in any of the jointtranslations. Mean anterior tibial translation curves werenearly identical for the intact and reconstructed limbs (seeFigure 6C), with a mean difference across all subjects andtime points of only 0.5 mm (reconstructed less than intact;P = .75). Three of the subjects had consistently greateranterior tibial translation in the reconstructed limb (eg,Figure 4C), whereas 3 had less. The estimated ACL-graft-length data were more consistent, with the functional graftlength shorter than the functional ACL length at 1 or moretime points for all subjects (and at all time points for 4subjects), with a mean decrease of –0.9 ± 1.48 mm across alltime points and subjects. This trend is visible in the meancurve (Figure 6D) but was not statistically significant(P = .20).

Four of the 6 subjects had greater lateral tibial transla-tion in the reconstructed limbs. Although this trend wasapparent in the group mean curve (Figure 6A), it was notsignificant for this sample size (mean difference = 1.12mm, P = .12). No trends were evident in the proximal/-

distal translations (Figure 6B).Translation magnitudes varied considerably across sub-

jects, with between-subject standard deviations rangingfrom 2 to 6 mm. This was due in part to geometric differ-ences (eg, size) between the joints, creating baseline shiftsin the translation curves. Because bone shape and sizewere consistent across the 2 limbs of each subject, these

Figure 2. Coordinate system definitions. Anatomical land-marks were identified using surface reconstructions fromsubject-specific CT scans. Fixed transformations betweenmarker-based (“M” superscripts) and anatomical (“A” super-scripts) coordinate systems were also determined from CT.Translations were expressed in the anatomical coordinatesystem fixed to the tibia.

Vol. 32, No. 4, 2004 Abnormal Rotational Knee Motion During Running 979

geometric differences did not affect the paired statisticalanalysis.

DISCUSSION

The primary clinical goals for ACL reconstruction surgeryare to restore stability and regain function lost due to ACLinjury. The primary instability traditionally associatedwith ACL loss is excessive anterior translation of the tibiarelative to the femur. Thus, reconstructive surgery is gen-erally considered successful if anterior tibial translation in

the reconstructed knee is similar to that in the contralat-eral, uninjured knee and the patient is eventually able toresume a preinjury activity level. Based on these evalua-tion criteria, ACL reconstruction is generally successful.The above results show that anterior instability is elimi-nated by the procedure under dynamic, stressful loading aswell as during static (KT-1000 arthrometer) testing.

The results of this study show, however, that restorationof normal anterior/posterior motion patterns does not nec-essarily imply normal knee function. Consistent, statisti-cally significant differences were found between uninjuredand reconstructed knee motion for both internal/external

TABLE 1Subject Characteristicsa

CincinnatiInjury→ Surgery→ KT-1000 Knee Score

Subject Age Sex Graft Type Additional Procedures Surgery, mo Test, mm Difference, mo (4 scales)

1 48 Female Hamstring Partial lateral meniscectomy 18 6 1 40/40/60/402 44 Female BPTB 24 5 2 80/80/80/703 42 Female BPTB 24 4 1 90/60/80/804 40 Male Hamstring 24 12 –1 55/55/80/405 40 Female Hamstring Medial meniscal repair 2 4 2.5 75/60/80/606 24 Male Hamstring Partial medial meniscectomy 12 4 1.5 60/80/100/80

aGrafts were either quadrupled hamstring tendon with interference screw fixation or bone–patellar tendon–bone with interference screwfixation. KT-1000 arthrometer measurements were displacement of reconstructed side minus displacement of uninjured side, at 89 N force.The 4 reported values in the Cincinnati Knee Ligament Rating System (as described by Noyes et al29) are (1) sports activity level, (2) abil-ity to participate in sports, (3) functional assessment rating scale, and (4) rate of pain, swelling, giving way; are all based on a 0% to 100%scoring system (with 100% being maximum function/minimum pain). KT-1000 arthrometer scoring and functional assessments were per-formed within 2 weeks of kinematic testing.

Figure 3. Joint rotational kinematics, typical subject. The dashed line is the intact (uninjured) limb; the solid line is the ACL-reconstructed limb.

980 Tashman et al The American Journal of Sports Medicine

rotation and abduction/adduction. These differences weresmall in absolute terms, averaging only 4° external rota-tion and 3° adduction. However, these differences are largepercentages of the range of observed motion about theseaxes (51% for internal/external rotation and 270% for ab-duction /adduction relative to the average range of motionin the uninjured limbs).

Three-dimensional animations of recorded tibiofemoralkinematics (using bone surfaces rendered from subject-specific CT data) provided a qualitative assessment ofthese rotational changes on motion at the articulating sur-faces of the tibiofemoral joint. The shift toward adductiontypically resulted from an increase in lateral compartmentseparation as well as a decrease in medial compartmentseparation, whereas the external tibial rotation shifted thetibial contact area primarily in the lateral compartment. Ifthe medial compartment contact point was unchanged, asimple geometric analysis showed that a 4° increase inexternal rotation would shift the lateral compartment tibialplateau contact area anteriorly by approximately 3.5 mmin a typical-size joint. Similarly, assuming equal effects inboth compartments, a 3° adduction would create a 1.3-mmloss of medial compartment separation and a 1.3-mm in-crease in lateral compartment separation. Thus, thesemovement pattern shifts could significantly alter the loca-tion, pattern, and magnitude of stresses applied to bothcartilage and menisci.

Two other kinematic parameters—medial/lateral trans-lation and ACL-graft length—showed trends suggestingdifferences between the uninjured and reconstructedlimbs. Although these trends did not reach statistical sig-nificance, the possibility of true differences cannot be ruledout until more subjects are included in the analysis. Themedial/lateral translation data (which were closest to sig-nificance at P = .12) had a power of only 0.32 to detect theobserved difference of 1.12 mm. A post hoc power analysissuggested that a sample size of 25 would be required todetect a difference of this magnitude with a power of 0.8.

There is considerable support in the literature to suggestthat mechanical abnormalities are associated with OAdevelopment and progression, based on both human andanimal studies.5,12,13,33,34 Knee adduction has been specifi-cally linked to both higher incidence and faster progres-sion of knee OA.8 Development of OA in the ACL-deficientdog has been shown4,32 and was thought to be driven pri-marily by the large anterior tibial shift created by ACLloss. However, recent high-speed radiographic studies haveshown that changes in the patterns of joint contact werebetter predictors of OA development in this model than themagnitude of anterior/posterior instability.46 Although theclinical significance cannot be assessed from the data pre-sented, possible links between the rotational abnormalitiesdetected in ACL-reconstructed knees and increased OAincidence in this population should be investigated further.

Figure 4. Joint translational kinematics, typical subject. The dashed line is the intact (uninjured) limb; the solid line is the ACL-reconstructed limb.

Vol. 32, No. 4, 2004 Abnormal Rotational Knee Motion During Running 981

These rotational abnormalities in ACL-reconstructedknees have not been previously reported. However, as dis-cussed in the introduction above, the techniques previouslyemployed to measure dynamic in vivo knee kinematics (eg,skin markers) may not have been sufficiently reliable toidentify the small (3°-4°) differences found in this study.Static studies of knee stability were typically designed totest the behavior of the knee only near the limits of thepassive range of motion of the joint. The results aboveshow shifts in the “operating point” of the joint but no sig-nificant increase in the range of motion or any evidencethat the joint is operating at or near the limits of passivestability. This may explain why traditional static stabilitymeasures have not been predictive of outcome.3,9,17,39,43

The combination of a somewhat stressful task, high-accuracy skeletal kinematics (free of skin motion artifact),

subject-specific 3D bone models, and within-subject side-to-side comparisons was sufficient to detect subtle differ-ences in dynamic joint behavior. The low across-trial stan-dard deviations (<3° across subject; typically <1° withinsubject) are evidence to the reliability of this technique.Although higher across-subject variability was observed inthe translational measurements (SDs ranging from 2 to6 mm), it was due primarily to anatomical differencesbetween the subjects. Because translations were measuredrelative to specific anatomical locations (estimated ACLorigin and insertion), larger joints would naturally tend tohave larger translation baselines (while rotations would beunaffected by joint size). Normalization of translations tojoint size might have reduced this variability but wouldnot have changed the statistical outcome (which was basedon paired within-subject comparisons). Within-subjecttranslation variability was much lower (typically less than1 mm).

A significant limitation of this study was the hetero-geneity of the subject population. Although no effects ofage, graft type, meniscal injury, surgery/testing timing, orfunctional scoring were detected, the sample size and sta-tistical power were probably insufficient to detect suchrelationships. Graft positioning was also not tightly con-trolled (although a coronal angle of 30° from vertical wastargeted). These factors could have a significant influenceon some of the results reported here, especially consideringthe relatively small sample size. For example, even thoughno significant group differences were found, translationaldifferences were identified in some subjects. This suggeststhat there may be additional trends within the data, butthere were insufficient subjects to link these observationsto any specific subgroup. Data on all these elements havebeen collected (including graft positioning), from CT. Asthis study is ongoing, it should be possible to investigatethe influence of these confounding factors as data frommore subjects become available.

It is important to note, however, that rotational abnor-malities reported above were consistent in both directionand magnitude even across this relatively diverse subjectgroup. In all subjects, reconstruction failed to restore nor-mal 3D rotational tibiofemoral motion under dynamic,stressful loading. Further research is required to deter-mine the relationship between these residual abnormali-ties and early-onset knee OA and to investigate modifica-tions to the reconstruction procedure that could reduceharmful movement abnormalities and improve long-termoutcome after ACL reconstruction (eg, changes in tunnelplacement, anatomical/double-bundle grafts).

ACKNOWLEDGMENT

This work was supported by a grant from the NationalInstitutes of Health, NIAMS AR46387.

REFERENCES

1. Alexander EJ, Andriacchi TP. Correcting for deformation in skin-based marker systems. J Biomech. 2001;34:355-362.

Figure 5. Joint rotational kinematics, average of 6 subjects.The dashed line is the intact (uninjured) limb; the solid line isthe ACL-reconstructed limb. Vertical lines are ±1 SD.

982 Tashman et al The American Journal of Sports Medicine

2. Andriacchi T, Toney M. A point cluster method for in vivo measure-ment of limb segment movement. Adv Bioengineer. 1994;28:185-186.

3. Barber SD, Noyes FR, Mangine RE, et al. Quantitative assessment offunctional limitations in normal and anterior cruciate ligament–deficient knees. Clin Orthop. 1990;255:204-214.

4. Brandt KD, Myers SL, Burr D, et al. Osteoarthritic changes in caninearticular cartilage, subchondral bone, and synovium 54 months aftertransection of the anterior cruciate ligament. Arthritis Rheum.1991;34:1560-1570.

5. Burr DB, Radin EL. Trauma as a factor in the initiation of osteoarthri-tis. In: Brandt KD, ed. Cartilage Changes in Osteoarthritis.Indianapolis: Indiana University School of Medicine; 1990:73-80.

6. Cappozzo A, Canati F, Leardini A, et al. Position and orientation inspace of bones during movement: experimental artifacts. ClinBiomech (Bristol, Avon). 1996;11:90-100.

7. Cappozzo A, Capello A, Della Croce U, et al. Surface marker clusterdesign criteria for 3-D bone movement reconstruction. IEEE TransBiomed Eng. 1997;44:1165-1174.

8. Cerejo R, Dunlop DD, Cahue S, et al. The influence of alignment onrisk of knee osteoarthritis progression according to baseline stage ofdisease. Arthritis Rheum. 2002;46:2632-2636.

9. Cross MJ, Wootton JR, Bokor DJ. Acute repair of injury to the anteri-or cruciate ligament: a long-term followup. Am J Sports Med.1993;21:128-131.

10. Daniel DM, Stone ML. KT-1000 anterior-posterior displacementmeasurements. In: Daniel DM, Akeson WH, O’Conner JJ, eds. Knee

Ligaments: Structure, Function, Injury, and Repair. New York, NY:Raven Press; 1990:427-447.

11. Daniel DM, Stone ML, Dobson BE, et al. Fate of the ACL-injuredpatient: a prospective outcome study. Am J Sports Med.1994;22:632-644.

12. Dekel S, Weissman SL. Joint changes after overuse and peak over-loading of rabbit knees in vivo. Acta Orthop Scand. 1978;49:519-528.

13. Felson D, Lawrence R, Dieppe P, et al. Osteoarthritis: new insights.Part 1: the disease and its risk factors. Ann Intern Med.2000;133:635-646.

14. Fetto JF, Marshall JL. The natural history and diagnosis of anteriorcruciate ligament insufficiency. Clin Orthop. 1980;147:29-38.

15. Fink C, Hoser C, Benedetto KP. Development of arthrosis after rup-ture of the anterior cruciate ligament: a comparison of surgical andconservative therapy [in German]. Unfallchirurg. 1994;97:357-361.

16. Grood ES, Suntay WJ. A joint coordinate system for the clinicaldescription of three-dimensional motions: application to the knee. JBiomech Eng. 1983;105:136-144.

17. Harter RA, Osternig LR, Singer KM. Long-term evaluation of knee sta-bility and function following surgical reconstruction for anterior cruci-ate ligament insufficiency. Am J Sports Med. 1988;16:434-443.

18. Holden JP, Orsindi JA, Siegel KL, et al. Surface movement errors inshank kinematics and knee kinetics during gait. Gait Posture.1997;5:217-227.

19. Karrholm J. Roentgen stereophotogrammetry: review of orthopedicapplications. Acta Orthop Scand. 1989;60:491-503.

Figure 6. Joint translational kinematics, average of 6 subjects. The dashed line is the intact (uninjured) limb; the solid line is theACL-reconstructed limb. Vertical lines are ±1 SD.

Vol. 32, No. 4, 2004 Abnormal Rotational Knee Motion During Running 983

20. Karrholm J, Elmquist LG, Selvic G, et al. Chronic anterolateral insta-bility of the knee: a roentgen stereophotogrammetric evaluation. AmJ Sports Med. 1989;17:555-563.

21. Karrholm J, Selvik G, Elmqvist L-G, et al. Three-dimensional instabil-ity of the anterior cruciate deficient knee. J Bone Joint Surg Br.1988;70:777-783.

22. Kuster M, Wood G, Sakurai S, et al. Downhill walking: a stressful taskfor the anterior cruciate ligament? A biomechanical study with clinicalimplications. Knee Surg Sports Traumatol Arthrosc. 1994;2:2-7.

23. Lafortune MA. The Use of Intra-Cortical Pins to Measure the Motionof the Knee Joint During Walking [PhD thesis]. State College:Pennsylvania State University; 1984.

24. Lafortune MA, Lambert CE, Lake MJ. Skin marker displacement atthe knee joint. In: NACOB II: The Second North American Congresson Biomechanics (J Biomech. vol 26). Chicago, Ill: Pergamon Press;1992:299.

25. Lynch MA, Henning CA. Osteoarthritis and the ACL-deficient knee. In:Feagin JA, ed. The Crucial Ligaments: Diagnosis and Treatment ofLigamentous Injuries About the Knee. New York, NY: ChurchillLivingstone; 1988:385-391.

26. Markolf KL, Kochan A, Amstutz HC. Measurement of knee stiffnessand laxity in patients with documented absence of the anterior cruci-ate ligament. J Bone Joint Surg Am. 1984;66:242-253.

27. Markolf KL, William BL, Shoemaker SC, et al. The role of joint load inknee stability. J Bone Joint Surg Am. 1981;63:570-585.

28. McDaniel WJ, Dameron TB. Untreated ruptures of the anterior cruci-ate ligament. J Bone Joint Surg Am. 1980;62:696-705.

29. Noyes FR, Barber SD, Mooar LA. A rationale for assessing sportsactivity levels and limitations in knee disorders. Clin Orthop.1989;246:238-249.

30. Noyes FR, Mooar PA, Matthews DS, et al. The symptomatic anteriorcruciate–deficient knee. J Bone Joint Surg Am. 1983;65:154-174.

31. O’Connor P, Oates K, Gardner DL, et al. Low temperature and con-ventional scanning electron microscopic observations of dog femoralcondylar cartilage surface after anterior cruciate ligament division.Ann Rheum Dis. 1985;44:321-327.

32. Pond MJ, Nuki G. Experimentally induced osteoarthrosis in the dog.Ann Rheum Dis. 1973;32:387-388.

33. Radin EL, Ehrlich MG, Chernack R, et al. Effect of repetitive impulsiveloading on the knee joints of rabbits. Clin Orthop. 1978;131:288-293.

34. Radin EL, Schaffler MB, Gibson GJ, et al. Osteoarthrosis as a resultof repetitive trauma. In: Osteoarthritis. Chicago, Ill: American Societyof Orthopaedic Surgeons; 1994.

35. Reinschmidt C, van den Bogert AJ, Nigg BM, et al. Effect of skinmovement on the analysis of skeletal knee joint motion during run-ning. J Biomech. 1997;30:729-732.

36. Rhoad RC, Klimkiewicz JJ, Williams GR, et al. A new in vivo tech-nique for three-dimensional shoulder kinematics analysis. SkeletalRadiol. 1998;27:92-97.

37. Schipplein OD, Andriacchi TP. Interaction between active and passiveknee stabilizers during level walking. J Orthop Res. 1991;9:113-119.

38. Selvik G. Roentgen stereophotogrammetric analysis. Acta Radiol.1990;31:113-126.

39. Seto JL, Orofino AS, Morrissey MC. Assessment of quadriceps/ham-string strength, knee ligament stability, functional and sports activitylevels five years after anterior cruciate ligament reconstruction. Am JSports Med. 1988;16:170-180.

40. Sgaglione NA, Del Pizzo W, Fox JM, et al. Critical analysis of knee lig-ament rating systems. Am J Sports Med. 1995;23:660-667.

41. Sheehan FT, Zajac FE, Drace JE. Using cine phase contrast magnet-ic resonance imaging to non-invasively study in vivo knee dynamics.J Biomech. 1998;31:21-26.

42. Shino K, Inoue M, Horibe S, et al. Measurement of the anterior stabil-ity of the knee: a new apparatus for clinical testing. J Bone Joint SurgBr. 1987;69:608-613.

43. Snyder-Mackler L, Fitzgerald GK, Bartolozzi AR, et al. The relation-ship between passive joint laxity and functional outcome after anteri-or cruciate ligament injury. Am J Sports Med. 1997;25:191-195.

44. Spoor CW, Veldpaus FE. Rigid body motion calculated from spacialcoordinates of markers. J Biomech. 1980;13:391-393.

45. Tashman S, Anderst W. In vivo measurement of dynamic joint motionusing high speed biplane radiography and CT: application to canineACL deficiency. J Biomech Eng. 2003;125:238-245.

46. Tashman S, Anderst W. Predicting cartilage damage from knee kine-matics in ACL-deficient dogs. Paper presented at: 47th AnnualMeeting of the Orthopaedic Research Society; 2001; San Francisco,Calif: 660.

47. Tegner Y, Lysholm J, Lysholm M. A performance test to monitor reha-bilitation and evaluate anterior cruciate ligament injuries. Am J SportsMed. 1986;14:156-159.

48. Torzilli PA, Greenberg RL, Hood RW, et al. Measurement of anterior-posterior motion of the knee in injured patients using a biomechani-cal stress technique. J Bone Joint Surg Am. 1984;66:1438-1442.

49. Winter DA. Processing of raw kinematic data. In: Winter DA, ed.Biomechanics of Human Movement. New York, NY: John Wiley;1990:24-39.

50. Youmans WT. The so-called “isolated” anterior cruciate ligament tearor anterior cruciate ligament syndrome: a report of 32 cases withsome observation on treatment and its effect on results. Am J SportsMed. 1978;6:26-30.